首页> 中文期刊>软件学报 >异构分布式环境下多DAG工作流的混合调度策略

异构分布式环境下多DAG工作流的混合调度策略

     

摘要

关于多个DAG工作流在异构分布式环境下调度的研究近来有了新的进展,也解决了一些问题,但现阶段还没有考虑和解决根据不同类型DAG的需求按优先级进行分类,以及对不同时间到达的多个不同优先级DAG进行调度的问题.为解决这些问题,针对各用户对DAG工作流的QoS需求的不同,在对不同用户的DAG工作流进行优先级划分的基础上,首先提出了一种新的调度模型,并改进了已有的公平调度算法,解决在不同时间上被提交的具有相同优先级的多个DAG工作流之间调度的公平性问题.为了提高资源利用率和高优先级DAG尽可能小地受低优先级DAG的影响,又提出了一种适用于多个不同优先级DAG之间调度的Backfill算法.在新的系统模型和这两种算法的基础上,提出了一种混合调度策略.实验结果表明,这种混合调策略能够兼顾不同时间到达的多个不同类型DAG调度需求和资源利用率的改善.另外,通过实验发现了关于两个DAG调度所特有的“拖尾”规律,具有进一步研 究和应用的价值.%Recent research in multiple DAG workflows in heterogeneous systems have been making progress and have solved some problems, but fail to classify the multiple DAGs, according to the demand of the performance asked by the varied DAG workflow and also fail to address the scheduling multiple DAGs workflow with multiple priorities submitted at different times. To solve these problems, the paper presents a new model of multiple DAGs management system for multiple DAGs workflow with multiple priorities and an adjustment method to the previous Fairness algorithm to solve the fairness issue in scheduling multiple DAGs with the same priorities submitted at different times. In addition, the study also proposes an implementation method of the Backfill algorithm for multiple DAGs with different priorities to improve utilization rate of resource, and then, based on the new model and the two methods, propose a hybrid strategy for scheduling multiple DAGs with multiple priorities submitted at different times. These experimental results show that it is possible to meet different requirements of DAGs submitted at different times and to improve utilization rate of a resource. In addition, the results about scheduling two-DAGs show a significant "Trail Ending" principle, which is valuable for academic study and application.

著录项

相似文献

  • 中文文献
  • 外文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号